Protein-ligand complexes
 Biological relevance
 Molecular recognition
 Structure of complexes
 Protein druggability
 Small molecules
 Molecular docking
 Evaluation of complexes
 Transport of small molecules
Outline
2Protein-ligand complexes
Protein-ligand complexes
3Biological relevance
Why do we care?
Examples?
 Cell signaling & regulation
 Binding of small molecules to receptors
 Molecular function of ligands/receptors
 Selectivity of receptors
 Signaling pathways
 Transport mechanisms
 Homeostasis of the cell
 …
Biological relevance
4Biological relevance
 Metabolism
 Binding of small molecules to enzymes
 Molecular function of enzymes
 Activation of enzymes and molecular pathways
 Bioactivation and clearance of drugs and xenobiotics (P450s,…)
 Enzymatic cascades
 Metabolic interferences (competing pathways)
 …
5Biological relevance
Biological relevance
 Drug discovery
 Binding of small molecules to macromolecules
 Identification of targets (enzymes, receptors, ...)
 Identification of potential target inhibitors/activators
 Optimization of target modulators
 Repurposing of drugs – finding new receptors
 Adverse side-effects due to binding
to off-targets
 …
6Biological relevance
Biological relevance
 Binding
 Specific binding governed by complementarity
 Geometry and shape
 Physicochemical properties (interactions)
Biophysical aspects
Histone octamer
Molecular recognition – biological roles 7
 Catalysis
 Chemical reactions can be accelerated up to 17 orders of
magnitude
 Binding to active site decreases the energy barrier of the reaction
 Stabilization of the Transition State(s)
Hammerhead ribozyme
Molecular recognition – biological roles
Biophysical aspects
8
 Signaling
 Conformational changes in response to
 Ligand binding
 Properties of surrounding environment (pH, forces… )
 Different conformations recognized by different proteins in
signaling pathways  control of cellular processes
Guanine riboswitch
Molecular recognition – biological roles
Biophysical aspects
9
 Formation of complex structures
 Structural elements of complex systems
 Governed by specific association of protein subunits
 With themselves
 Other proteins, carbohydrates, lipids, …
Molecular recognition – biological roles
Biophysical aspects
10
 Molecular recognition refers to the specific interactions
between two or more molecules through non-covalent bonding
 Different biological roles
 Specific binding
 Catalysis
 Signaling
 Several models to explain molecular recognition
Molecular recognition
Molecular recognition 11
 E. Fisher – 1894
Lock-and-key model
Molecular recognition – mechanisms 12
What is it?
 E. Fisher – 1894
 Complementarity between receptor’s binding site
and the ligand
 Size & shape
 Physicochemical properties
 Both ligand and receptor are considered rigid
 Not sufficient to explain allostery, non-competitive inhibition,
or catalysis
  Model dismissed, only used for educational purposes
Lock-and-key model
Molecular recognition – mechanisms 13
 D. E. Koshland – 1956
Induced-fit model
Molecular recognition – mechanisms 14
What is it?
 D. E. Koshland – 1956
 Only partial complementarity necessary
 Both ligand and receptor can undergo conformational
adjustments upon complexation
 Conformation of the bound receptor does not exist in its free state
Induced-fit model
Molecular recognition – mechanisms 15
 B. F. Straub – 1964
 This model is also called: conformational selection, fluctuation-fit
or population selection
 Receptor and ligand flexible  considered as ensembles
 Complex is formed in a lock-and-key fashion when two
complementary configurations occur
 Conformation of the bound receptor exists also in its free state
Selected-fit model
Molecular recognition – mechanisms 16
 Z. Prokop – 2012
 When the receptor has a buried active site and tunnels
 Complementarity with the ligand is needed both for the
active site and the tunnel
 Explains the extra selectivity filter provided by the tunnel
Keyhole-lock-key model
Molecular recognition – mechanisms 17
reactants
products
 Enzymes increase the speed of chemical reactions by
decreasing the activation barrier
Biocatalysis
Molecular recognition – biocatalysis
▪ Kinetic rate:
(Arrhenius equation)
▪ Lower Ea  higher k
(faster reaction)
Ea
Ea
H
with
enzyme
without
enzyme
G‡
enzyme-substrate
complex enzyme-products
complex
transition
state
𝑘 = 𝐴𝑒
−𝐸 𝑎
𝑅𝑇
18
transition
state
 Enzymes increase the speed of chemical reactions by
decreasing the activation barrier
 Provide environments that stabilize the transition state(s)
Biocatalysis
Molecular recognition – biocatalysis
Ea
Ea
H
with
enzyme
without
enzyme
G‡
reactants
enzyme-substrate
complex enzyme-products
complex
products
transition
state
19
Structures of complexes
20Structure of complexes
 Complexes in RSCB PDB
 Databases of complexes
 PDBbind
 BindingDB
 ChEMBL
 …
 Experimentally determined complexes!
Complexes in RSCB PDB
21Structure of complexes
 Limited number of available complexes
 >180,000 protein structures
 >101,000 structures with ligands
 Limited information on conformation of bound ligand
 Ligands are often mobile  uncertainties  need to be verified
Databases of complexes
22Structure of complexes
 PDBbind
 http://www.pdbbind.org.cn
 Curated binding affinity data and structural information on >16,500
complexes
 >13,500 protein-ligand
 >120 nucleic acid-ligand
 >800 protein-nucleic acid
 >2,000 protein-protein complexes
 Data collected from >29,000 original references
 Provides also a "refined set" and "core set" compiled as high-quality
data sets of protein-ligand complexes for docking/scoring studies
Databases of complexes
23Structure of complexes
 PDBbind
Databases of complexes
24Structure of complexes
 PDBbind
Databases of complexes
25Structure of complexes
 PDBbind
Databases of complexes
26Structure of complexes
 PDBbind
Databases of complexes
27Structure of complexes
 BindingDB
 www.bindingdb.org
 The first public molecular recognition database
 Focused on the interactions of proteins considered to be
drug-targets with drug-like molecules
 Contains about 1,500,000 entries of binding data
 >7,000 protein targets
 >650,000 small molecules
 Crystal structures of complexes with measured affinity
 >2,500 – for proteins with 100% sequence identity
 >6,000 – for proteins up to 85% sequence identity
Databases of complexes
28Structure of complexes
 BindingDB
Databases of complexes
29Structure of complexes
 BindingDB
Databases of complexes
30Structure of complexes
 ChEMBL
 https://www.ebi.ac.uk/chembldb/
 Is a manually curated database of bioactive molecules with
drug-like properties
 Database of binding, functional and ADME (Absorption,
Distribution, Metabolism, and Excretion) and toxic. information
 Contains >15,000,000 activity data
 >12,000 protein targets
 >1,700,000 distinct small molecules
 Data collected from >67,000 original publications
 Smart clustering of relevant information
Databases of complexes
31Structure of complexes
 ChEMBL
Databases of complexes
32Structure of complexes
 ChEMBL
Databases of complexes
33Structure of complexes
 ChEMBL
 Druggability
 Likelihood of a particular protein to be modulated or targeted by a
drug-like molecule in a way that leads to a therapeutic effect
 Meaning, it can bind with high affinity to selective, bioavailable,
low-molecular weight molecules
 Lipinski’s rule of 5 (for orally-active drugs)
 MW ≤ 500 Da
 ≤ 5 H-bond donors (NH, OH); ≤ 10 H-bond acceptors (F, O, N)
 Partition coefficient (log Po/w) ≤ 5
 Usually 1 violation is acceptable
Protein druggability
34Protein druggability
Protein druggability
35Protein druggability
 Druggability
Protein druggability
36Protein druggability
 Druggability
 Prediction of protein druggability
 By similarity to known target
 Sequence of binding domain
 Structural features of binding sites
 From databases of known targets
 Predictive tools: PockDrug Server, DoGSiteScorer, …
 Important in target identification phase of drug discovery
 Unfortunately, many resources are only private or
commercial
Protein druggability
37Protein druggability
 PockDrug-Server
 http://pockdrug.rpbs.univ-paris-diderot.fr/
 Automatic tool combining pocket detection, characterization and
druggability prediction
 Based on:
 Physicochemical features
 Geometry, volume, shape
 Druggability probability for
one pocket or two
pockets for comparison
Protein druggability server
38Protein druggability
 Proteins Plus
 https://proteins.plus/
 Meta-server providing global support for the initial steps in
analysing protein structures
 Structure search, quality assessment, protein pocket detection,
protein-ligand and protein-protein
interactions
 Predicts binding sites and estimates
their druggability (using DoGSiteScorer)
Protein-ligand interactions server
39Protein druggability
Break time
 Representation of small molecules
 Databases of small molecule
 Cambridge Structural Database
 PubChem database
 ZINC database
 Preparation of small molecule structure
Small molecules
41Small molecules
 1D – atom based (empirical formula)
 C2H5Cl
 2D – chemical structure diagram
 Topology or SMILES (Simplified Molecular Line Entry System)
 3D – atomic coordinates
 Usually: PDB, SDF or MOL2 files
 Beware: may have different protonation states
Representation of small molecules
42Small molecules
CCCl C1=CC=C(C=C1)CN
Databases of small molecule
 Cambridge Structural Database
 http://www.ccdc.cam.ac.uk/products/csd/
 The world largest repository of crystal structures of small molecules
 >900,000 curated & validates structures with experimental
3D coordinates available
 CSD is distributed commercially
 Free interactive demo for educational purposes
(only ~750 structures)
 https://www.ccdc.cam.ac.uk/Community/educationalresources/
teaching-database/
Small molecules 43
Databases of small molecule
 Cambridge Structural Database
Small molecules 44
Databases of small molecule
 PubChem
 http://pubchem.ncbi.nlm.nih.gov/
 World largest open repository of experimental data identifying the
biological activities of small molecules
 Substances: >270 M chemical entities
 Compound: >111 M unique chemical structures. Compounds may
be searched by chemical properties and are pre-clustered by
structure comparison into identity and similarity groups
 BioAssays: >1.4 M biological experiments
 Bioactivities: >300 M biological activity data points
Small molecules 45
Databases of small molecule
 ZINC database
 http://zinc.docking.org/
 Free public resource for ligand discovery
 3D coordinates in ready-to-dock formats (ex: added hydrogens,
partial atomic charges, … )
 Molecules in biologically relevant protonation and tautomeric forms
 About 37 billion unique molecules grouped by classes
 >750,000,000 – commercially available molecules
 >10,000,000 – drug-like molecules
 > 5,000 – FDA-approved drugs
 …
Small molecules 46
Preparation of small molecule structure
 AVOGARO
 https://avogadro.cc/
 Free, open-source molecule editor and visualizer
 Intuitive & easy to use
 Useful to convert file formats
 Embedded molecular minimization and molecular mechanics
 Interface to quantum chemistry packages
Small molecules 47
Preparation of small molecule structure
 AVOGARO
Small molecules 48
Preparation of small molecule structure
 PyMOL
 https://pymol.org/
 Powerful molecular visualizer and editor
Small molecules 49
Preparation of small molecule structure
 Open Babel
 https://openbabel.org/
 Free, open-source
 Widely used molecule format converter
 Command line and graphical interface
Small molecules 50
Molecular docking
51Molecular docking
What is it?
 Useful when experimental data is not available
or for virtual screening
Molecular docking
52Molecular docking
Crystal
(experimental)
Docking attempts
Score
RMSD
 Several components/steps
 Receptor representation
 Ligand representation
 Search of binding modes
 Scoring
Molecular docking
53Molecular docking
Receptor Ligand Complex
 Receptor represented only by relevant binding site
 Descriptor representation – derived from geometry
and interaction abilities of binding site
(H-bond donor/acceptor, hydrophobic contacts, …)
 Grid representation – entire search region
is covered by orthogonal equidistant
points carrying information about chemical
properties or the interactions of probe
atom at those points with the receptor atoms
Precomputing properties can speed up calculations
Receptor representation
54Molecular docking – receptor
 Receptor flexibility
 Fully rigid approximation
 Soft docking – employs tolerant “soft” scoring functions to simulate
plasticity of otherwise rigid receptor
 Explicit side-chain flexibility – optimization of residues by rotating
part of their structure or rotation of whole side-chains using
predefined rotamer libraries
 Docking to molecular ensemble of protein structure – obtained
from multiple crystal structures, from NMR structure determination
or from a trajectory produced by MD simulation
Receptor representation
55Molecular docking – receptor
 Ligands represented by all atoms or just some
 Non-polar hydrogens can be united with their respective parent
carbon atoms to reduce number of atoms in calculation
 Ligand flexibility
 Only rotation about single bonds
 Docking of a library of pre-generated ligand conformations –
applicable only to quite rigid ligands due to exponential increase in
number of possible conformers with number of rotatable bonds
 Direct sampling of ligand conformational space during searching
 Fragment-based techniques – ligand is cut into several fragments
and rigidly docked into binding site
Ligand representation
56Molecular docking – ligand
Molecular docking – search
57Molecular docking – search
 Many search algorithms available
 Rigid docking 
 Semi-flexible 
 Fully flexible 
(but demanding)



 Geometry-based and combinatorial algorithms
 Assumes that binding is governed by shape and/or physicochemical
complementarity between the ligand and the receptor
 Assumes that the degree of complementarity is proportional to
the binding energy which is not always true especially for
more polar ligands
 Energy-driven and stochastic algorithms
 Tries to locate directly the global minimum of the binding free
energy corresponding to the experimental structure
 Random basis of these methods requires multiple independent runs
of docking calculations to achieve consistent results
Molecular docking – search
58Molecular docking – search
 Matching algorithms
 Represent a ligand and a receptor binding site by descriptors
derived from their geometry and/or presence of particular
interaction sites
 Try to align/match complementary parts of ligand and binding site
and in this way predict the ligand binding mode
 SW packages
 DOCK – http://dock.compbio.ucsf.edu/
 SLIDE – http://www.kuhnlab.bmb.msu.edu/software/slide/
 …
Geometry-based algorithms
59Molecular docking – search
 Matching algorithms
Geometry-based algorithms
60Molecular docking – search
 Fragment-based algorithms
 Ligand is initially fragmented into rigid parts
 Two approaches to obtain whole docked molecule
 Incremental construction – fragments are incrementally
docked into the receptor until whole ligand is constructed
 Fragment-placing and linking – all fragments are docked
simultaneously and then joined together
 SW packages
 FlexX – http://www.biosolveit.de/FlexX/
 eHITS – http://www.simbiosys.ca/ehits/
 …
Geometry-based algorithms
61Molecular docking – search
 Fragment-based algorithms
Geometry-based algorithms
62Molecular docking – search
 Monte Carlo algorithms
 Explore protein-ligand interactions space by iteratively introducing
random changes into a position, orientation or conformation of the
ligand and evaluating new configuration using acceptance criterion
 New configuration is always accepted if its energy is more favorable
then the energy of previous configuration or accepted with some
probability reflecting energy difference to previous configuration
 SW packages
 Autodock Vina – http://vina.scripps.edu
 Glide – http://www.schrodinger.com/Glide
 …
Stochastic energy-driven algorithms
63Molecular docking – search
 Monte Carlo algorithms
Stochastic energy-driven algorithms
64Molecular docking – search
 Genetic algorithms
 Configurations of the ligand from randomly generated initial
population are encoded in their “genes” which are subject of
random genetic modification (single point mutation, crossover, …)
 Individuals with better fitness (binding energy) have higher
chance to survive and reproduce to next generation
 Overall fitness of population is increasing with each new generation
 SW packages
 AutoDock – http://autodock.scripps.edu
 GOLD – http://www.ccdc.cam.ac.uk/products/life_sciences/gold/
 …
Stochastic energy-driven algorithms
65Molecular docking – search
 Genetic algorithms
Stochastic energy-driven algorithms
66Molecular docking – search
 Scoring function
 Evaluate all the binding modes from the searching algorithms
 Must be computationally efficient and provide accurate description
of protein-ligand interactions
 Application of scoring functions to rank
 Several configurations of one ligand bound to one protein –
essential for prediction of the best binding mode
 Different ligands bound to one protein – determination of
substrate or inhibitor specificity
 One ligand bound to several different proteins – functional
annotation of proteins and study of drug selectivity
Molecular docking – scoring
67Molecular docking – scoring
 Categories of scoring functions
 Empirical
 Knowledge-based
 Force field-based
 Machine learning
Molecular docking – scoring
68Molecular docking – scoring
 Categories of scoring functions
 Empirical
 Derived by fitting the following equation to experimental
binding affinities of known protein-ligand complexes
 Rapid evaluations
 Arbitrary selection of terms included in the equation  failure
when binding is governed by any excluded type of interaction
 Weights are dependent on the chosen training set
Molecular docking – scoring
69Molecular docking – scoring
.......  rotellipohbbind GGGGG 
 Categories of scoring function
 Knowledge-based
 Capture the knowledge about protein-ligand binding that is
implicitly stored in structural data by statistical analysis
 Atom-pair potentials derived from
distances found for such pair
in training structural data
 Rapid evaluations
 Describe all types of interactions without any preselection
 Problem when structural data do not contain sufficient
information on specific atom-pairs (ex. halogens, metals, …)
Molecular docking – scoring
70Molecular docking – scoring
 Categories of scoring function
 Force field-based
 Use the non-bonded terms from well-established force fields
 Provide precise affinities
 Computationally demanding  employed for rescoring
selected binding modes (not during searching)
Molecular docking – scoring
71Molecular docking – scoring
 Intermolecular interactions
 Binding energies
Evaluation of complexes
72Evaluation of complexes
 Most common types
 Hydrogen bonds
 Hydrophobic
 Aromatic
 Ionic interactions
Intermolecular interactions
73Evaluation of complexes
 Visualization
 Schematic diagrams showing hydrogen bonds and hydrophobic
contacts
 Tools
 LigPlot+
 Stand alone application
 http://www.ebi.ac.uk/thornton-srv/software/LigPlus/
 Pre-calculated for protein-ligand complexes in PDBsum
(pictorial database of PDB structures)
Intermolecular interactions
74Evaluation of complexes
 Binding Affinity Prediction of Protein-Ligand (BAPPL) server
 http://www.scfbio-iitd.res.in/software/drugdesign/bappl.jsp
 Calculates binding free energy
of a protein-ligand complex
using all-atom-energy-based
empirical scoring function
 Only for non-metallo protein-ligand
complexes
Binding energies
75Evaluation of complexes
 Describe trajectory of ligands through tunnels
 Based on geometry w/wo molecular docking
 Fast but low accuracy
 Good for screening purposes
 CaverDock, MoMA-LigPath, SLITHER
 Based on force field
 Run multiple MD simulations
 Accurate but computationally demanding
 Metadynamics, steered MD, adaptive sampling, etc.
Transport of small molecules
76Transport of small molecules
 CaverDock
 https://loschmidt.chemi.muni.cz/caverdock/
 Analysis of tunnels by Caver
 Discretization of identified tunnel into discs
 Molecular docking by AutoDock Vina to every disc
 Caver Web
 https://loschmidt.chemi.muni.cz/caverweb/
 Web interface for Caver and CaverDock
Transport of small molecules
77Transport of small molecules
CAVER
Discretization
CaverDock
78
CaverDock
Transport of small molecules
Active
site
 Results provided:
 Ligand trajectory
 Energy profile
79
CaverDock
Transport of small molecules
-5
-3
-1
1
3
5
7
9
11
0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
2.1
2.3
2.5
0 5 10 15
Energy(kcal/mol)
Tunnelradius(Å)
Trajectory along the tunnel [Å]
+
Tunnel
bottleneck
80
CaverDock over Caver Web
Transport of small molecules
References I
 Gu, J. & Bourne, P. E. (2009). Structural Bioinformatics, 2nd Edition,
Wiley-Blackwell, Hoboken.
 Pérot, S. et al. (2010). Druggable pockets and binding site centric
chemical space: a paradigm shift in drug discovery. Drug Discovery
Today 15: 656-667.
 Moitessier, N. et al. (2008). Towards the development of universal, fast
and highly accurate docking/scoring methods: a long way to go. British
Journal of Pharmacology 153: S7-S26.
References 81
References II
 Bolton, E. E. et al. (2008). PubChem: Integrated platform of small
molecules and biological activities. Annual Reports in Computational
Chemistry 4: 217-241.
 Gaulton, A. et al. (2012). ChEMBL: a large-scale bioactivity database for
drug discovery. Nucleic Acids Research 40: D1100-D1107.
 Irwin, J. J. et al. (2012). ZINC: A free tool to discover chemistry for
biology. Journal of Chemical Information and Modeling 52: 1757-1768.
 Santos, R. et al. (2017). A comprehensive map of molecular drug targets.
Nature Reviews Drug Discovery. 16: 19-34
References 82